Order books vs AMMs in prediction markets
Prediction markets use one of two fundamentally different mechanisms to match buyers and sellers: a traditional orderbook, or an automated market maker (AMM). The choice shapes spreads, liquidity, and what kinds of markets can exist at all.
Order books vs AMMs in prediction markets
Orderbooks: match-when-they-meet
Kalshi and Polymarket both use orderbook models. Buyers post limit orders ("I'll buy YES at $0.55 or below"), sellers post limit orders ("I'll sell YES at $0.57 or above"), and trades execute when a buy price meets or exceeds a sell price. In liquid markets this produces tight spreads — often one or two cents — and excellent price discovery. In illiquid markets the orderbook can be empty on one side, and you simply cannot trade at any price.
AMMs: algorithmic quotes
Automated market makers replace the orderbook with a smart-contract liquidity pool that prices shares using a mathematical formula. The most famous is Robin Hanson's LMSR (Logarithmic Market Scoring Rule), which always provides a quote, even in dead markets, and bounds the market maker's maximum loss. The trade-off: AMMs give worse prices on large trades because the curve moves against you with size.
Which is better?
For high-volume markets — elections, major sports, macro data — orderbooks win on spread and execution quality. For long-tail questions with few traders, AMMs are the only way a market can exist at all. Many modern platforms use hybrid approaches: an orderbook for the main flow with an AMM filling in when the book is thin.
Related in this cluster
What is implied probability in prediction markets?
Why a YES share trading at $0.62 means a ~62% market-implied probability — and when that reading breaks down.
ClusterWhat is LMSR (Logarithmic Market Scoring Rule)?
The market-maker formula Robin Hanson designed for prediction markets, why it bounds loss, and who still uses it.
ClusterHow liquidity works in prediction markets
Why liquidity is the single biggest predictor of accuracy, how thin markets are manipulated, and what "depth" really means.